Exploring syntactic categories in a construction
grammar framework
Christina Galeano
Simon Fraser University
Within a construction grammar framework, evidence is provided
for the idea that syntactic structures are constructions. Syntactic
categories such as NP and VP are posited to yield higher
representations of meaning beyond the simple sum of their
constituents, which fits the general description of a construction.
NPs and VPs are analyzed as contributing a conceptual meaning
of “grounding”, in accord with Langacker’s grounding theory.
Furthermore, IP represents a grounded proposition, which is
ultimately utilized by the more complex “question construction”
via conceptual inheritance. Conceptual inheritance between
deep and surface structure is explored for question constructions.
The ultimate idea is that base-generated and “derived” syntactic
categories are not related by pure structural movement. Rather,
their relation stems from common conceptual representations,
with additional concept representations for the derived category.
1
Central ideas
Over the last few years, there has been renewed interest in a cognitive linguistic
framework for defining human language, as set out by researchers like Langacker
and Goldberg, to name a couple. The central theme behind the cognitive
linguistic framework is that every linguistic form correlates to some sort of
meaning. This idea is particularly relevant to the traditional domain of syntax,
where we might wonder whether syntactic categories can be considered linguistic
forms as well. By synthesizing the work of the above-mentioned authors, I will
show that syntactic categories are indeed forms, as they render meaning
correlates. In particular, I will discuss how syntactic categories are a particular
type of form known as a construction, following the work of Adele Goldberg’s
Construction Grammar theory. I will use Goldberg’s theory in conjunction with
Langacker’s theory of grounding to offer an alternative account for so-called
“derived” syntactic categories (such as those categories found in a questions,
thought to arise from movement). It is my hope that this analysis will aid in the
overall goal of defining a “cognitive grammar”.
112
2
What is a construction?
In her theory of construction grammar, Adele Goldberg posits that all linguistic
forms are constructions (Goldberg, 1995). Since, in a cognitive linguistic
framework, all forms correlate to some sort of construed meaning, constructions
correlate to meaning. “Constructions” can be taken to mean any salient form
where some sort of representation of meaning is construed. Examples of different
types of constructions are shown in the figure below.
Construction
Form/Example
Morpheme
Anti-, pre-, -ing
Word
Avocado, anaconda, and
Complex word
Daredevil, shoo-in
Idiom-partially filled
Jog (someone’s) memory
Ditransitive Construction
Sub [V Obj1 Obj2]
Function
Transfer
Figure 1. Representatives of different construction types. Function is
labeled where it is not immediately identifiable. (Goldberg 220)
As can be seen from the figure, constructions vary in size and complexity,
where even words and morphemes are considered constructions. The important
thing to note here is that the meaning of the construction comes from the
construction itself as a whole. For example, Goldberg argues that the
“ditransitive construction” above provides a meaning of “transfer” as a whole.
This construction (V Obj1 Obj2) is what is responsible for our ability to
understand both simple sentences like “pass me a wrench” alongside sentences
like “cry me a river”. The idea here is that both sentences employ a ditransitive
construction that contains the meaning of “transfer.” As such, metaphors and
idioms that employ this construction can also be understood due to the
construction alone, and not just by the individual lexical items contained within.
Another important aspect of the constructions is that each one yields a
meaning representation that is more than just the sum of its constituent elements,
or more than any inherent properties of the element itself. The “synergistic
Working Papers of the Linguistics Circle of the University of Victoria 21, 111–117
© 2011 Christina Galeano
113
effect” of constructions is exemplified by Langacker, who asks us to consider
three words: “boy” “girl”, and “like” (Langacker, 1991). If we just had these
words in isolation-- “boy girl like”-- there could be a number of different
meanings. It could mean “the boy likes the girl”, “the girl might like the boy”,
“the girl liked the boy”, etc. However, in arranging these in a particular
configuration, say “the boy like the girl” 1 (along with adding articles), we gain a
greater meaning, which is more than just the sum of the elements in question.
How does this work? For starters, we get the notion of a NP when we can
combine “girl” with “the” in the correct order. The same goes for “boy”.
Similarly, when the word “like” is followed by a NP, a VP results.
3
Can syntactic categories be considered constructions?
From Langacker’s example, we see that the syntactic categories NP and VP are
salient forms that have some sort of construed meaning. Furthermore, NP and VP
both denote more abstract representations than just the sum of their constituent
elements. These representations cannot be predicted from the constituents in any
way. This idea begs the question of whether syntactic categories can also be
considered constructions as well, and subsequently, what meaning they might
correlate to.
The main question we ask is what meaning is construed by NP and VP
constructions. Though the answer to this question is not certain, Langacker
posits that a process called grounding yields these categories. Grounding is the
process of fixing conceptual image schemas in place and time for actions
(Langacker, 1987). Therefore, the categories NP and VP provide additional
meaning of “entity or action grounded in space or time” when they are construed.
To see clearly how this works, take “boy”. “Boy” is not grounded until it is put
with an article—a, the, etc2. Grounding, and the NP category that results, thus
serves the function of yielding a particular type of representational schema for
this entity. Similarly “like” alone does not acquire the higher category
representation of “verb: action” until it’s placed into some sort of proper
grounding construction, such as between a subject and an object. If it were not
for constructions that ground linguistic elements, we would get a verbal
representation schema from ‘like’ in this construction as well: ‘the boy the girl
1
The inflectional suffix –s is left out because it is considered to be agreement that
expresses redundant quantity information about the subject. I will address redundancies
later.
2
In languages that do not have articles, the noun itself comes with a representation for
groundedness “built in”, which English does not have. For a more detailed explanation of
definiteness/indefinite articles and how they ground, see Langacker and Fauconnier.
Working Papers of the Linguistics Circle of the University of Victoria 21, 111–117
© 2011 Christina Galeano
114
like’. However, native English speakers do not get a clear verbal representation,
as this particular construction doesn’t exist in English to ground ‘like’ in time.
Therefore, it is the construction itself that yields the higher representational
meaning, and this representation is greater than just the sum of its parts.
From the construed meaning of the NP and VP categories, how do we
arrive at the total meaning given by “the boy like the girl”? Langacker does not
address the notion of the sentence itself as a whole being a syntactic category
construction. In combining NP and VP, we get a higher representational notion
of yet another category construction. This construction is traditionally designated
as IP, or the sentential level of representation. In construing the entire sentence
construction “the boy likes the girl”, we inevitably end up with a notion like
‘directional grounded event (like) from one grounded participant to another.”
This meaning cannot be predicted from the constituent NP and VP themselves,
nor from the semantics of the word-level constructions beneath them, so it is
worthwhile to consider this “propositional/IP level” as a construction too.
4
Constructions all the way down
Goldberg argues that human language grammars can be defined by an
interconnected network of constructions from various levels. Larger
constructions can contain smaller, simpler constructions for meaning, such as IP
containing NP and VP within it. The idea that larger constructions can house
smaller ones is what Goldberg terms “constructions all the way down”
(Goldberg, 2003). As an example, she shows the multilayered and embedded
meaning constructions involved in the sentence “What did Liza buy the child”.
1
1. Words
2. Ditransitive construction :
3. Question construction :
4. Subject-Auxiliary inversion
construction:
5. VP construction
6. NP constuction
2
[ What
----
3
did
4
Liza
5
6
7
buy the child? ]
--------------------
-----------
----
----
------------------------------
Figure 2. An expression, or ‘construct’, that is a combination of constructions
shown in row 1, coded (with dashes) to the appropriate parts of the expression
(VP, Verb-Phrase; NP, Noun-Phrase) (Goldberg, 2003, p. 221).
Here, we see an example of various levels of constructions interconnected
with one another. At the lowest level, there are words/morphemes (which are
themselves constructions). When the words are concatenated into certain
Working Papers of the Linguistics Circle of the University of Victoria 21, 111–117
© 2011 Christina Galeano
115
configurations, such as “the child”, another representation manifests, one that we
typically designate by NP, so Goldberg calls it an NP construction. From
Langacker’s example, we saw that the representation of NP usually indicates a
meaning of groundedness of an entity. Therefore, we consider “Liza” and
“what” to be inherently grounded, whereas the grounding of “child” happens
when it is concatenated with “the”.
In construing the different lexical items [Liza, buy, the, child, what, did],
there are a number of different constructions that we can make beyond just the
NP and the VP constructions. For example, NPs gain the additional meaning of
“object” when placed into a “ditransitive construction”: V Obj1 Obj2—“Buy the
child what”, in this case. The whole ditransitive construction gives a meaning of
“transfer”. Still further, there is a “question construction”, taken from the whole
utterance with “what” at the beginning. This too is considered a construction, as
it builds an additional layer of meaning—the notion of a question. Larger
constructions inherit the properties of smaller ones. That is, the overall question
construction in the example above inherits the properties of the smaller ones
within, such as the subject-auxiliary construction, the NP construction, the VP
construction, etc.
It is important to reiterate here that the meaning of a construction comes
from the construction itself. Smaller constructions combine in a certain
meaningful way to yield a new construction. Their combination results in a new
representational meaning layer beyond just the sum of the lower constructions it
contains.
5
Accounting for Syntactic Categories thought to Arise from Movement:
Differing Conceptual Complexity
We saw that constructions can be embedded within constructions. However,
Goldberg makes the claim that the two sentences “Liza did buy the child what”
and “What did Liza buy the child” are two different constructions (Goldberg
2003). This idea opposes the generative theory, where one is structurally derived
from the other through movement. In the generative framework, the syntactic
category CP is thought be created for questions when we move from the deep
structure (Liza did buy the child what) to a surface structure for the question
(what did Liza buy the child). The idea is that we access the base structure first,
and then go from there to the target structure by moving key elements. However,
Goldberg’s model dictates that different constructions, no matter how complex
they are and how much they seem to be based off of one another, are basegenerated in situ. That means that “What did Liza buy the child” and “Liza did
buy the child what” are both considered to be base-generated exactly as their
surface structures appear.
At first glance, it is difficult to think that these two constructions are not
Working Papers of the Linguistics Circle of the University of Victoria 21, 111–117
© 2011 Christina Galeano
116
related to one another, which is what makes the generative rule of movement
from a deep to a surface structure so attractive. However, Goldberg’s theory
implies that these are two separate constructions, despite how similar they are. If
they are not derived from purely structural movement, what else could be
responsible for their similarity? Though Goldberg says these are two different
constructions, it doesn’t necessarily entail that they are two unrelated
constructions. The difference between her theory and generative grammar
theories, though, is that these constructions are considered to be related via
conceptual derivation, rather than being related through pure structural
derivation. With conceptual derivation, the meaning in one construction is
inherited by the meaning of a larger encompassing construction, as we saw
before. For example, the overall question construction “What did Liza buy the
child” inherits the meaning of the constructions contained within it, like the
subject-auxiliary construction and the ditransitive construction.
Let’s consider how conceptual derivation would look in IP and question
constructions, given our ideas about IP grounding. Recall the claim that
IP/propositional constructions represent a grounded situation, with grounded
entities and actions in space and time. However, when we employ a question
construction, we construe an additional layer of meaning to the original
proposition. The additional meaning that is construed is that of making part of
the construction virtual or ungrounded (symbolized by ‘what’ in this case). The
conceptual inheritance of constructions is more salient in the example “do you
swing dance”, where the grounded proposition (you swing dance) is made virtual
in space and time by being used in the CP question construction3. In other words,
the question construction has the effect of “ungrounding” the entire
proposition/IP.
In cases like the IP and question constructions above (“Liza did buy the
child what” and “What did Liza buy the child”) Goldbeg says these are two
different overall constructions. However, they employ many of the same smaller
constituents. Rather than thinking that this comes about through movement from
the deep structure, we can offer a conceptual inheritance interpretation of the
similarities instead. If we construe one situation as a basic proposition, we use
the IP construction. If, however, we construe a basic proposition, along with it
not being grounded in space/time, we use the question construction (which
inherits the IP construction). In other words, two constructions can be related
conceptually (where one is the basic proposition, and the other is the proposition
plus the added meaning of “virtual/ungrounded” element), which is enough to
account for their similarities.
3
This ability to establish spaces where events and participants are not necessarily in the
“here and now” is a crucial element to displacement, which gives rise to a number of
syntactic categories, and lends incredible expressive power to human language.
Working Papers of the Linguistics Circle of the University of Victoria 21, 111–117
© 2011 Christina Galeano
117
6
Conclusion
In a cognitive linguistic framework, we define all linguistic forms as
constructions, which correlate to some sort of construed meaning. Under this
approach, the grammar of a language is defined by interconnected constructions
for meaning that occur at many different levels (“constructions all the way down”
according to Goldberg). The question was whether syntactic categories can be
considered constructions as well, and how they are construed and represented.
We saw evidence for the idea that syntactic categories like NP, VP, and IP are
constructions, as they reflect the multilayered construal of meaning just as other
constructions do. Syntactic categories that are considered to be structurally
derived, such as those used in question formation, actually represent complex
constructions that inherit the meaning of lower-level constructions. Thus,
Goldberg’s hypothesis of different constructions for seemingly-related sentences
is not necessarily disproven by the existence of conceptual inheritance. In
investigating how syntactic categories are construed and constructed, we get at
broader questions of whether syntax is autonomous, or whether it is driven by
general cognitive construals of meaning, such that we can consider syntactic
categories to be constructions.
References
Carey, S. (2009). The origin of concepts. New York: Oxford University Press.
Fauconnier, G, & Turner, M. (1994). Conceptual projection and middle
spaces. Cognitive Science Technical Report, 9401. Retrieved from
http://www-cogsci.ucsd.edu/index.php?cat=research&page=publicationstech_report
Goldberg, A. (2003). Constructions: a new theoretical approach to
language. TRENDS in Cognitive Sciences, 7(5), 219-224.
Goldberg, A. (1995). Constructions: a construction grammar approach to
argument structure. Chicago: University of Chicago Press
Langacker, R. (1991). Foundations of cognitive grammar. Stanford: Stanford
University Press.
Langacker, R. (1987). Nouns and verbs. Language, 63(1), 53-94.
Working Papers of the Linguistics Circle of the University of Victoria 21, 111–117
© 2011 Christina Galeano